SearchQA
收藏资源简介:
该数据集名为SearchQA,它基于一系列《危险边缘》节目的线索构建而成,包含了复杂的问题以及谷歌搜索结果前几条的相关片段。每个线索都与正确答案及一系列片段相关联,且用于答案跨度选择的环境仅限于前10个片段。该数据集规模宏大,包含了超过14万个问题/答案对和690万个片段;具体分为训练集99,820个例子,验证集13,393个例子,以及测试集27,248个例子。其任务是进行问题回答。
The dataset named SearchQA is constructed from clues sourced from a series of *Jeopardy!* episodes. It encompasses complex questions and relevant snippets extracted from the top-ranked results of Google searches. Each clue is associated with its correct answer and a set of snippets, with the context for answer span selection limited to the top 10 snippets. This is a large-scale dataset containing over 140,000 question-answer pairs and 6.9 million snippets. Specifically, it is split into 99,820 training examples, 13,393 validation examples, and 27,248 test examples. The core task of this dataset is question answering.

- 1SearchQA: A New Q&A Dataset Augmented with Context from a Search Engine纽约大学数据科学中心 · 2017年



